Skip to content
#

ycb-objects

Here are 7 public repositories matching this topic...

Language: All
Filter by language
sim2real-domain-randomization

How much visual variety does a simulator need before a model trained in it works on real photographs? Six ResNet-18 classifiers, 150,000 synthetic renders, 3,000 real YCB test images, zero real training data. Randomization takes accuracy from 8.6% to 61.3%, and Grad-CAM shows why: attention moves from background to object.

  • Updated Sep 9, 2026
  • Jupyter Notebook

Add this topic to your repo

To associate your repository with the ycb-objects topic, visit your repo's landing page and select "manage topics."

Learn more